Automatic Content Generation in Tetris Game Based on Emotion Modeling

Experiencing the personalization via the player's modeling as well as appropriate adjustment of the content according to specific needs and preferences was important steps towards effective content generation. For the experiments presented here, a version of the classic Tetris game is enhanced with 4 controllable parameters. We extracted 57 features derived from gameplay interaction, which are probably associated with emotions such as fun, challenge and frustration. The pairwise preference data of the player was collected by using forced choice questionnaires, and the models were trained by using 5 machine learning algorithms after the selection of the features.